Skip navigation links
Saturday, November 21st 2009 02:01:01 PM PST
tcc-2-72.local
tcc-3-71.local
The Triton Resource Data Oasis is designed as an extremely large scale storage system, having two-to-four petabytes of total disk capacity when fully deployed. The system will provide between 60 and 120 gigabytes per second of data movement bandwidth and manage from 3000 to 6000 individual disks. Data Oasis is now in the planning stage. More information will be provided here soon.
There are currently three data storage systems on Triton Resource:
Home areas on Triton are located on Solaris-based NFS servers using ZFS as the underlying file system.
Triton has a parallel file system. The first revision has approximately 100 terabytes (approximately 115 trillion bytes) of space. The second revision will be called Data Oasis and is expected to deliver 8 times the aggregate bandwidth and 10 times the available space.
A small scratch space is available to all users
Each system has different performance characteristics, backup policies, hardware, and data management software. See the table below for a summary, or view the Backup Policy page for complete details.
The current Lustre Storage Area is a Parallel File System (PFS) called Mirage. It contains about 100 terabytes of shared scratch space available to all users. When completed, the Data Oasis system will replace Mirage and become the primary Lustre PFS for Triton.
This PFS is fully connected to both the Petascale Data Analysis Facility and the Triton Compute Cluster, providing adequate data movement and data management throughput to users on either system. The scratch space currently available will be the resource capacity until the completion of Data Oasis.
In addition to the shared PFS storage, there are 36 terabytes of space in the Home Storage system for all users. This will be the maximum size for the home filesystem for the foreseeable future. There are also approximately 5 gigabytes per node of Local Temporary Storage available during job runs. This space is purged between jobs.
| System | Local HD Redundancy | Connectivity | Space | Backup/Replication | Total Space for files, snapshots | Location |
|---|---|---|---|---|---|---|
| Home Area Storage | Double-drive failure can occur without data loss: Triton uses Raidz2 variant of Raid6 | 10GbE; Delivers > 300MB/sec to single node; > 500MB/sec aggregate | Each user is guaranteed at least 50GB of space | Nightly snapshots retained 7 days minimum; view the snapshot recovery page |
36TB | Snapshots accessible at $HOME/.zfs/snapshot |
| Lustre Storage PFS (Mirage) | Single-drive hardware failure is supported through Raid5 on the Lustre Object | 4 x 10GbE; delivers > 500MB/sec to single node; > 2.5GB/sec aggregate | min/max currently undefined; Triton supports project special requests |
No backup of this storage is performed | 100TB | Storage accessible at /mirage/<username> |
| Local Node Temporary Space | Single-drive hardware failure is supported with Linux SW Raid-1 (Mirroring) | Local HD; about 50MB/sec/node; about 14GB/sec aggregate | Generally about 5GB/node; purged between jobs | No backup of this storage is performed | Dependent on number of nodes requested for job | Storage accessible at /tmp from local node only |
View a rack diagram of the complete Triton Resource.
View a complete system diagram of the Triton Resource.
Download the Gigabit Ethernet Network Diagram (PDF).
Open a Ticket with Triton Resource Support using the Support Ticket Form.
Join the Discussion Forum Sign up for our Email Discussion List.
FAQ Read the FAQ Page.
